Thinking Machines, an AI startup founded by former OpenAI CTO Mira Murati (미라 무라티), said on May 11 it has unveiled a research preview of its “Interaction Model,” an AI model designed to handle real-time interaction by default.

Thinking Machines pointed to what it called a problem with current AI models: they operate in a turn-based mode, generating a response only after a user finishes input. In this mode, a model cannot take in new information while producing an answer, making it hard for a user to stay in the loop.

The Interaction Model, by contrast, was designed from the start with a focus on real-time responsiveness. The company said it adopts a multi-stream, micro-turn approach to process audio, video and text at the same time, while thinking, responding and acting in real time.

The Interaction Model can implicitly determine whether a user is speaking, thinking or waiting for a reply, allowing it to continue a conversation naturally without a separate dialogue management component. It can step in in context even if a user has not finished speaking, and the user and the model can talk at the same time, enabling tasks such as real-time interpretation. The company also said it can handle web search, browsing and UI generation while talking with a user, and integrate the results naturally into the conversation.

Thinking Machines said it built interaction capabilities as part of the model itself, unlike existing AI models that implement such functions by connecting external components. It stressed that this improves collaboration as the model becomes smarter.